Efficient Coupling of Light to Graphene Plasmons by Compressing

Apr 28, 2014 - Here, we propose and numerically demonstrate coupling between infrared photons and graphene plasmons by the compression of surface pola...
1 downloads 16 Views 3MB Size
Letter pubs.acs.org/NanoLett

Efficient Coupling of Light to Graphene Plasmons by Compressing Surface Polaritons with Tapered Bulk Materials A. Yu. Nikitin,†,‡ P. Alonso-González,† and R. Hillenbrand*,†,‡ †

CIC nanoGUNE Consolider, 20018 DonostiaSan Sebastian, Spain IKERBASQUE Basque Foundation for Science, 48011 Bilbao, Spain



S Supporting Information *

ABSTRACT: Graphene plasmons promise exciting nanophotonic and optoelectronic applications. Owing to their extremely short wavelengths, however, the efficient coupling of photons to propagating graphene plasmonscritical for the development of future devicescan be challenging. Here, we propose and numerically demonstrate coupling between infrared photons and graphene plasmons by the compression of surface polaritons on tapered bulk slabs of both polar and doped semiconductor materials. Propagation of surface phonon polaritons (in SiC) and surface plasmon polaritons (in n-GaAs) along the tapered slabs compresses the polariton wavelengths from several micrometers to around 200 nm, which perfectly matches the wavelengths of graphene plasmons. The proposed coupling device allows for a 25% conversion of the incident energy into graphene plasmons and, therefore, could become an efficient route toward graphene plasmon circuitry. KEYWORDS: graphene plasmons, surface polaritons, plasmon compression, mode coupling

G

order of the wavelength of the incident light) is required, which in turn requires a large propagation length of the GPs and prevents compactness of the device. Another classical geometry common for the excitation of surface plasmons is the prism coupling method.16 Taking into account the large GP momenta at mid-IR and THz frequencies, the refractive index of the prism would need to be very large (∼10−100), and thus is practically not available. We could also think of the excitation of GPs by compression of the photon wavelength in a dielectric waveguide with gradually increasing refractive index n = n(x) and subsequent waveguide coupling to GPs, analogous to coupling plasmons in metal films to dielectric waveguides.17,18 However, compression of the mode wavelength in dielectric waveguides to the scale of GPs would again require materials with refractive indices up to 10 to 100 (as in the case of a prism coupler), which do not exist in nature at optical and infrared frequencies. Metamaterials could be used instead, which have been shown to support optical modes with ultrahigh refractive indices.19 In this case, however, one faces an extremely challenging and highly demanding fabrication process. Here, we introduce a novel concept (illustrated in Figure 1) for the efficient launching of graphene plasmons based on the compression of surface polaritons (SPs) on tapered slabs of a bulk (3D) material supporting either surface phonon polaritons

raphene plasmons (GPs) are electromagnetic waves propagating along a graphene layer.1−10 Their electrostatic tunability and extremely short wavelength, λp, being much smaller than the corresponding photon wavelength in free space, λ0, promise exciting optoelectronic applications at the nanoscale. At infrared frequencies, GP wavelengths as short as 200 nm have been already observed experimentally.8,9 Because of their extremely high momentum, GPs strongly concentrate electromagnetic energy, promising novel nanoscale photonic applications such as ultracompact tunable plasmonic absorbers,11,12 sensors, waveguides or modulators.10,13 However, the huge momentum mismatch between free space photons and GPs challenges the efficient coupling between them. Recently, it has been reported that propagating GPs can be excited by the strongly concentrated fields at the apex of metallic near-field probes.8,9 The efficiency of the excitation mechanism, however, is unknown and expected to be low. Further, the experimental configuration of near-field microscopy does not constitute a practical solution for the development of integrated GP devices. The design of efficient couplers for GPs still remains a challenging task. Several other configurations to overcome the momentum mismatch between the incoming light and GPs have been studied theoretically. For example, a periodic grating (placed for instance on top of a dielectric waveguide) with sufficiently short period Λ can compensate the mismatch due to Bragg scattering in different diffraction orders n following the expression kp = k0 + nG, with G = 2π/Λ.14,15 For efficient excitation of GPs, however, a large coupling length (in the © 2014 American Chemical Society

Received: March 12, 2014 Revised: April 8, 2014 Published: April 28, 2014 2896

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901

Nano Letters

Letter

higher momenta and are more sensitive to the slab thickness (compared to the symmetric solutions). For the substrate, we choose ε2 = 2 (i.e., a low refractive index material) and for the upper half space, ε1 = 1 (air). In Figure 2a we show the dispersion for SiC slabs of different thickness d. It was obtained by solving numerically eq 1. The

Figure 1. Schematic of the proposed coupling device.

(SPhPs, for example, on SiC slabs) or surface plasmon polaritons (SPPs, for example on highly doped n-GaAs slabs). The coupling mechanism is as follows: first, an incident propagating light beam is coupled via total internal reflection to an SP on a relatively thick slab (Otto prism coupling20,21).22 The thickness of the slab is chosen such that the SP wavelength can be easily matched by the photon wavelength in the prism, enabling high coupling efficiencies.20,21 Subsequent propagation of the SP along a tapered slab yields a compression of its wavelength. The slab thickness is gradually reduced until the SP wavelength matches that of the GPs. Finally, by placing the graphene above the thinned slab, the field of the SP couples efficiently to the GP. We note that SPP compression on metal tapers has been already demonstrated at visible frequencies.23−26 However, at mid-IR and THz frequencies, well below the plasmon resonance frequency, the compression of electromagnetic energy does not provide the necessary wavelength reduction, even if the mode diameter is compressed to below 100 nm.27,28 In order to achieve wavelength compression, the dielectric permittivity of the taper material, εm, needs to satisfy Re(εm) < −1 (to support surface waves) and |Re(εm)| ≫ |Im(εm)|, with |Re(εm)| being small (to provide sufficiently high momenta and low losses). Within the wide range of exciting plasmonic materials,29 highly doped semiconductors such as n-GaAs30 and polar crystals such as SiC31−33 fulfill these requirements in the infrared and mid-infrared spectral range, where GPs currently attract much interest.10,11 We first consider the dispersion relation of SPs in thin slabs of constant thickness d. For a slab with dielectric permittivity εm embedded between two dielectric half-spaces with permittivities ε1 and ε2, the dispersion relation k = k(ω) follows from the solution of21 q qz1ε2 + qz 2ε1 i tan(qzmk 0d) = zm · 2 εm qz1qz 2 + qzm (ε1ε2 /εm2) (1)

Figure 2. Dispersion of the surface polaritons in slabs placed on a substrate and the dispersion of the graphene plasmons. (a) Slabs of SiC. The Fermi level of graphene is |EF| = 0.44 eV. (c) Slabs of nGaAs. The carrier concentration of the semiconductor is 1020 cm−3, the rest of the parameters are taken from.29 The Fermi level of graphene is |EF| = 0.8 eV. b,d) Snapshots of the vertical electric field Re(Ez) for different thicknesses of the slabs d. For comparison, the dispersion curves for the semi-infinite polaritonic media bounding either to air or to the substrate are shown in Figure 2a and c (solid and dashed red curves). The refractive index of the substrate is ε2 = 2 in all cases.

dielectric function of SiC was taken from Palik,34 and is given by ⎛ ⎞ ω2 − ω 2 εm = ε∞⎜1 + 2 LO 2 TO ⎟ ωTO − ω − iωγ ⎠ ⎝

(2)

with ε∞ = 6.56 the background permittivity, and ωLO, ωTO, and γ = 5.9 cm−1 the transversal and longitudinal phonon frequencies and relaxation rate, respectively. Between the transversal and longitudinal phonon frequencies, ωTO = 800 and ωLO = 970 cm−1, the dielectric function of SiC is negative, which indicates that SiC slabs support SPhP in this spectral range. For all thicknesses d, we see the typical dispersion for SPhPs. The wavevector k increases with frequency ω, reaching a maximum at approximately the surface phonon resonance at Re(εm) = −1 (ω = 950 cm−1). Notice that a double-humped behavior near phonon resonance is due to nonsymmetric dielectric surrounding of the film. At each hump the mode is

where k0 = ω/c is the momentum in free space, qz,1,2,m = (ε1,2,m − q2)1/2, q = k/k0, and Im(qz,1,2,m) ≥ 0. Equation 1 can be derived from the Helmholtz equation, taking into account the boundary conditions for the fields at the interface between the slab and the dielectric half spaces. The solution of eq 1 yields symmetric and antisymmetric modes with respect to the zcomponent of the electric field Ez.21 In the following, we will consider only the antisymmetric solutions, as they exhibit 2897

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901

Nano Letters

Letter

Figure 3. Snapshots of the real part of the vertical component of the electric field Re(Ez) and the spatial distribution of the absolute value of the electric field |E| for the modes propagating along the tapered waveguides. (a,b) SiC waveguide with an initial and a final thickness of 400 and 50 nm, respectively; the tapering angle is 4 deg. The frequency is 889 cm−1. (c,d) n-GaAs waveguide with an initial and a final thickness of 100 and 5 nm, respectively; the tapering angle is 2.7 deg. The frequency is 2016 cm−1. The fields are normalized to the maximum value of |E| of the incoming mode. The refractive index of the substrate is ε2 = 2, as in Figure 2.

localized at one of the film faces (compare the frequency positions of the humps in the dispersion curves for the semiinfinite SiC bounding either to air or to the substrate). For slabs thicker than 300 nm, the dispersion is similar to that of SPhPs on bulk SiC surfaces, with wavevectors k being not more than a factor of 2 to 5 larger than the corresponding photon wavevector k0. With decreasing d, we find a dramatic increase of the SPhP wavevector, which can be 10 to 50 times larger than k0. Figure 2b shows the mode profiles (real part of the electric field) for ω = 890 cm−1 (indicated by a dashed light gray curve in Figure 2a), illustrating the decreasing SPhP wavelength with decreasing thickness d. For slabs as thick as 50 nm, the SPhP wavelength is already 20 times smaller than the photon wavelength. Analogous to SPPs on metals, the SPhP wavelength reduction is accompanied not only by a huge field confinement, as is clearly observed in Figure 2b, but also by strong damping. The propagation length, however, is still in the order of several SPhP wavelengths. Most intriguing, the SPhPs wavevectors on 50 nm thick SiC slabs can be as large as the GPs wavevectors and between 890 and 920 cm−1 can even exceed them. We show this finding by plotting in Figure 2a the dispersion of GPs (dashed line, see details in the next paragraph) in a free-standing doped graphene layer. As an example we have chosen a Fermi energy |EF| = 0.44 eV, for which mid-infrared graphene plasmons have been already observed experimentally.8 The GP dispersion for free-standing graphene in the IR region, where the GP momentum is large, reads1 qGP ≃ i/α

with

α = 2πσ /c

SPs with extraordinarily high wavevectorssimilar to GPs can be also supported by thin slabs of a doped semiconductor material, as we demonstrate in Figure 2c,d for the case of ndoped GaAs with a carrier density of n = 1020 cm−3 and mobility of 103 cm2V−1s−1.29 The SPPs dispersion in doped semiconductors can be calculated following eq 1 with the dielectric function εm described by the Drude-Lorentz model εm = εb −

with

αeff = ik 0d

εm + ε1ε2 /εm ε1 + ε2

ω(ω + iγ )

(4)

with εb = 10.91 the background permittivity, and ωp = 11485 and γ = 137 cm−1 the plasma frequency and relaxation rate, respectively. The plasma frequency ωp ∼ √n is determined by the mobile carrier concentration n in the semiconductor, which can be adjusted by the degree of doping, thus providing a tuning capability for the SPP dispersion. Analogously to SPhPs on thin SiC slabs, we observe in Figure 2c,d SPPs on thin ndoped GaAs slabs, which large momenta reach well above the momenta of GPs. This can be appreciated by comparing the SPP dispersion with the GP dispersion on graphene with a Fermi energy of 0.8 eV (dashed line in Figure 2c). Altogether, Figure 2 clearly shows that thin slabs of a bulk material can support SPs with wavelengths comparable to that of GPs in the infrared and mid-infrared spectral regions. Although GPs may offer better tunability and simpler fabrication procedures, thin slabs of a bulk material offer the possibility to adjust the SP wavelength and mode confinement at a given frequency over a much larger range, from nearly the wavelength of incident photons to the much smaller wavelength of GPs and beyond. Such wavelength adjustments can be accomplished by simply reducing the slab thickness from a few hundreds of nanometers to a few tens of nanometers (actually, SPhPs with ultra-short wavelength have been recently verified by infrared nanoimaging of thin hBN layers37). In particular, the wavelength reduction offers the possibility to couple photons to GPs by propagating SPs along tapered slabs. In Figure 3, we demonstrate the compression of SPhPs and SPPs on tapered SiC (Figure 3a,b) and n-GaAs (Figure 3c,d) slabs, respectively. The frequency for the case of SiC is 889 cm−1, whereas for n-GaAs it is 2016 cm−1. We show the real part of

(2)

where σ is the optical conductivity calculated within the local random phase approximation.3,35,36 Interestingly, for large SPhP momenta and vanishing slab thickness (qdk0 ≪ 1), the dispersion relation 1 greatly simplifies to q ≃ i/αeff

ωp2

(3)

This shows that SPhPs in a very thin 3D SiC slab can be described similarly to GPs (2), by introducing an effective normalized conductivity αeff proportional to the thickness of the layer d. 2898

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901

Nano Letters

Letter

Figure 4. Snapshots of Re(Ez) and the spatial distribution of |E| for a tapered SiC waveguide coupled to a graphene sheet. The tapering angle is 15 deg, its length 1.3 μm. The initial and final thicknesses of the SiC slab are 400 and 50 nm, respectively. The vertical separation between graphene and SiC is 150 nm. (a,b) Infinite waveguide. (c,d) Cut waveguide. The length of the thinnest SiC part is 700 nm. The SiC-graphene overlap along the x axis is 600 nm. The frequency is 877 cm−1. The dashed arrows in (a,b) indicate the position of the field minimum in the SiC waveguide, while in (c,d) they show the termination of the SiC waveguide. The substrate and the parameters of graphene are the same as in Figure 2. The fields are normalized to the maximum value of |E| of the incoming mode. The upper schematics show the geometry of the structures. The insets show a zoomin to the taper region.

the vertical electric field along the tapered slab (thickness reduces from left to right) in Figure 3a,c and the corresponding absolute value of the field in Figure 3b,d. We clearly see that both the mode wavelength and mode volume decrease in the propagation direction (from left to right), which in turn induces a pronounced intensity enhancement at the end of the taper. For the SPhPs on the SiC taper, the mode index increases from qin = 2.62 + 0.14i (400 nm slab thickness) to qout = 17.4 + 1.25i (50 nm slab thickness), whereas for n-GaAs taper, we find qin = 1.76 + 0.069i (100 nm slab thickness) and qout = 21.2 + 2.26i (5 nm slab thickness). At the end of the taper, the SPs continue their propagation over several wavelengths along a 50 nm thick SiC slab and a 5 nm thick GaAs slab, respectively. However, the intensity decays significantly within a few mode wavelengths. Due to the subwavelength-scale field confinement, a large part of the electromagnetic energy propagates inside the slab, thus damping the mode. The mode propagation length in units of the SP wavelength (figure of merit) is given by L = Re(k)/ [2πIm(k)], yielding Lin = 2.98 at the beginning of the SiC taper (400 nm slab thickness) and Lout = 2.21 at its end (50 nm slab thickness). Remarkably, the strong mode compression along the taper does not significantly reduce L, highlighting that SiC is an interesting low-loss polaritonic material. Note that the strong reduction of the absolute propagation length is mainly due to the dramatic reduction of the mode wavelength. For the GaAs taper we find Lin = 4.06 (100 nm slab thickness) and Lout = 1.49 (5 nm slab thickness). In contrast to SiC, the propagation length relative to the mode wavelength diminishes clearly. We finally demonstrate in Figure 4 how the compression of SPs on tapered slabs can be employed for the efficient launching of GPs. We consider in the following a tapered SiC slab, because of the relatively weak losses during the mode compression. In order to achieve effective coupling to GPs, the graphene sheet is placed at a distance of 150 nm above the thin SiC slab extending from the taper. From the point of view of fabrication, this could be accomplished by depositing a thin dielectric layer onto the SiC slab (for instance CaF2 or SiO2) and then the graphene on top of this layer (see Figure 1). As

the optical properties of the dielectric layer will not strongly affect the coupling (provided that the dielectric permittivity of the layer εd is small), we consider in the following εd = 1, corresponding to free-standing graphene. In Figure 4a and b, we show colorplots for both the real part of the vertical field and the absolute value of the total field along the coupling device, Re(Ez) and |E|. The tapered SiC slab and parameters are the same as in Figure 3a,b. For the graphene layer, we assume a Fermi energy of |EF| = 0.44 eV, which is close to typical values observed in CVD grown graphene.8 For the mobility, we assume 11.36 × 103 cm2/(V·s) (corresponding to 0.5 ps relaxation time of the charge carriers). This value is elevated compared to graphene samples of recent plasmonic studies,8,9 but it allows for better illustration of the coupling between the slab SPs and GPs. On the other hand, such mobilities seem to be available in the future with graphene samples of improved quality.38,39 Figure 4a and b demonstrate the coupling between compressed surface phonon polaritons on a thin SiC slab and a graphene layer, both being infinitely long (right side of the figures). We observe that the field intensity (Figure 4b) is periodically transferred from the SiC slab to the graphene and vice versa. Such behavior is well known from waveguide couplers18,40−42 and originates from the beating between the two modes in closely spaced waveguides. In our case, the beating modes originate from the coupling of plasmons in the graphene sheet and the SPhPs in the SiC slab. According to Figure 4c, the beating period is Lb = 1.4 μm. This can be compared with the results of coupling mode theory,40 according to which Lb = 2π/|Re(k1 − k2)| where k1 and k2 are the wavevectors of the modes of the waveguide coupler. From the mode analysis (not shown here) of the coupled infinite graphene sheet and 50 nm thick SiC slab, we find k1/k0 = 19.22 + 0.39i and k2/k0 = 11.27 + 0.76i, so that Lb = 1.43 μm, which is in an excellent agreement with the value found from the full wave calculations (Figure 4b). Notice that the length at which the maximum energy transfer occurs is Lb/2. By terminating the SiC slab at the position of the maximum field in graphene (marked by arrows in Figure 4a and b), the 2899

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901

Nano Letters

Letter

of the SPhPs along the taper. Further reduction of the coupling efficiency might be caused during the coupling of photons to the initial SPhP mode on the thick slab. By proper design of a prism or grating coupler,21 photons might be converted with high efficiency into the initial SPhP mode on the 400 nm thick SiC slab, owing to its rather large wavelength of 5.25 μm (=0.46 λ0). In the following numerical study of the coupling efficiency, we therefore only consider the energy transfer from the initial SPhP mode to the GP mode in the configuration shown in Figure 4c and d. To that end, we define the coupling efficiency η = A/P0, where A is the energy absorbed by the graphene (dominated by dissipation of the GPs) and P0 the power flux at the beginning of the taper. In Figure 5a, we show the efficiency η as a function of the tapering angle θ for two different frequencies and different final thicknesses of the SiC slab: ω = 877 cm−1 and d2 = 50 nm (red curve); ω = 855 cm−1 and d2 = 25 nm (blue curve). The initial thickness of the slab is d1 = 400 in both cases. The length of the final slab, xc = 700 nm (which is approximately a half of a beating period xc ≃ Lb/2), has been chosen such that the termination of the slab coincides with the position of maximum energy transfer into the graphene layer (see Figure 4b). We find the maximum efficiency for taper angles θ = 15 and 11.5 degrees, respectively, which we explain by the compromise between field absorption in the taper (increasing with increasing taper length, i.e., smaller taper angle) and back reflections at and inside the taper (increasing with increasing taper angle, owing to an increasing impedance mismatch). In Figure 5b, we study how the efficiencies depend on the frequency. Each spectrum shows a resonance peak, which appears at the frequency where the SPhP and GP dispersion curves cross, that is, where the momentum of the SPhP best matches the GP momentum (see Figure 2a). Note that the curves in Figure 5a were calculated at the frequencies where the efficiency reaches its maximum (877 cm−1 for the red curve and ω = 855 cm−1 for blue curve, respectively). According to the results presented in Figure 5, the maximum efficiency of the coupler is higher than 25% at a frequency ω = 877 cm−1. This is a rather large value, especially when considering that the coupling length is comparable to the GP wavelength and that the SP mode is progressively absorbed during its propagation along the taper. To better understand the coupling losses, we evaluated the coupling efficiency independent of the taper (see Supporting Information). To that end, we calculated the coupling of SPhPs in an untapered 50 nm thick SiC slab (corresponding to the thinnest part of the taper) to GPs. We found a coupling efficiency of about 70% at ω = 877 cm−1. This means that the main amount of losses are due to the SPhP compression in the taper and not due to losses during the coupling process. In conclusion, we have demonstrated that by propagating surface phonon or plasmon polaritons along tapered slabs of a bulk (3D) polaritonic material, it is possible to efficiently launch graphene plasmons with a predicted efficiency of more than 25%. We further stress that thin slabs of bulk (3D) polaritonic materialsby themselvescould become an interesting platform for the development of short-wavelength surface polariton photonics for sensing and communication applications, particularly in the infrared spectral range where phonons and plasmons in materials such as SiC or n-GaAs exhibit low losses compared to plasmonic materials typically used at visible and near-IR frequencies (e.g., gold).

transferred SPhPs energy continues its propagation as plasmon in the graphene sheet (Figure 4c and d). In the presented example, the thin SiC slab is terminated at a distance xc ≃ Lb/2 = 700 nm from the taper end (see also schematics in Figure 5). We note that the coupling mechanism (after the mode compression) is similar to that of dielectric photonic modes and surface plasmons in metal slabs.17,18,41

Figure 5. Coupling efficiency of the cut tapered SiC waveguide. The initial thickness of the waveguide is d1 = 400 nm. The length of the thickest slab is 5 μm. The graphene vertical and horizontal separations are zg = 150 nm and xg = 100 nm, respectively. a) The efficiency as a function of the tapering angle θ for two different final slab thicknesses d2 and frequencies. The red curve is for ω = 877 cm−1 and d2 = 50 nm, whereas the blue curve is for ω = 855 cm−1 and d2 = 25 nm. b) The efficiency as a function of frequency for two different values of d2 and angles θ. The length of the thinnest part is xc = 700 nm in both configurations.

Figure 4c and d indicate already an efficient excitation of propagating GPs within a coupling length Lc = 600 nm, which is well below the photon wavelength (λ0 = 11.4 μm). This represent a significant advantage compared, for instance, to grating couplers on dielectric waveguides, where the coupling length needs to be larger than one photon wavelength.15 For that reason, even GPs with short propagation lengths can be efficiently excited. Due to the beating, the energy from the SPhP can be almost totally transferred to GPs (see below and Supporting Information). However, the total efficiency for converting photons into GPs is smaller, owing to losses (absorption, outof-plane scattering and back-reflection) during the compression 2900

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901

Nano Letters



Letter

(15) Gullans, M.; Chang, D. E.; Koppens, F. H. L.; de Abajo, F. J. G.; Lukin, M. D. Phys. Rev. Lett. 2013, 111 (24), 247401. (16) Bludov, Y. V.; Vasilevskiy, M. I.; Peres, N. M. R. Europhys. Lett. 2010, 92 (6), 68001. (17) Ditlbacher, H.; Galler, N.; Koller, D. M.; Hohenau, A.; Leitner, A.; Aussenegg, F. R.; Krenn, J. R. Opt. Express 2008, 16 (14), 10455− 10464. (18) Nesterov, M. L.; Kats, A. V.; Turitsyn, S. K. Opt. Express 2008, 16 (25), 20227−20240. (19) He, Y.; He, S.; Gao, J.; Yang, X. J. Opt. Soc. Am. B 2012, 29 (9), 2559−2566. (20) Raether, H. Surface Plasmons on Smooth and Rough Surfaces and on Gratings; Springer: Berlin/Heidelberg, 1988. (21) Agranovich, V. M.; Mills, D. L. Surface polaritons: electromagnetic waves at surfaces and interfaces; North-Holland Pub. Co.: Amsterdam, 1982. (22) In the schematic, the coupling of incident light to the polaritons in the slab is realized via Otto prism configuration, but other alternative excitation schemes are also possible (Kretchmann prism configuration, grating coupling, etc.). (23) Stockman, M. I. Phys. Rev. Lett. 2004, 93 (13), 137404. (24) Ropers, C.; Neacsu, C. C.; Elsaesser, T.; Albrecht, M.; Raschke, M. B.; Lienau, C. Nano Lett. 2007, 7 (9), 2784−2788. (25) Zaccaria, R. P.; De Angelis, F.; Toma, A.; Razzari, L.; Alabastri, A.; Das, G.; Liberale, C.; Di Fabrizio, E. Opt. Lett. 2012, 37 (4), 545− 547. (26) Gramotnev, D. K.; Bozhevolnyi, S. I. Nat. Photon 2014, 8 (1), 13−22. (27) Rusina, A.; Durach, M.; Nelson, K. A.; Stockman, M. I. Opt. Express 2008, 16 (23), 18576−18589. (28) Schnell, M.; Alonso Gonzalez, P.; Arzubiaga, L.; Casanova, F.; Hueso, L. E.; Chuvilin, A.; Hillenbrand, R. Nat. Photon 2011, 5 (5), 283−287. (29) Naik, G. V.; Shalaev, V. M.; Boltasseva, A. Adv. Mater. 2013, 25 (24), 3264−3294. (30) Wasserman, D.; Shaner, E. A.; Cederberg, J. G. Appl. Phys. Lett. 2007, 90 (19), 191102. (31) Greffet, J.-J.; Carminati, R.; Joulain, K.; Mulet, J.-P.; Mainguy, S.; Chen, Y. Nature 2002, 416 (6876), 61−64. (32) Taubner, T.; Korobkin, D.; Urzhumov, Y.; Shvets, G.; Hillenbrand, R. Science 2006, 313 (5793), 1595. (33) Borstel, G.; H. J. F Electromagnetic surface modes; Wiley: New York, 1982. (34) Palik, E. D. Handbook of Optical Constants of Solids; Elsevier Science: Cambridge, MA, 1985. (35) Hwang, E. H.; Das Sarma, S. Phys. Rev. B 2007, 75 (20), 205418. (36) Falkovsky, L. A. Phys.−Usp. 2008, 51 (9), 887−897. (37) Dai, S.; Fei, Z.; Ma, Q.; Rodin, A. S.; Wagner, M.; McLeod, A. S.; Liu, M. K.; Gannett, W.; Regan, W.; Watanabe, K.; Taniguchi, T.; Thiemens, M.; Dominguez, G.; Neto, A. H. C.; Zettl, A.; Keilmann, F.; Jarillo-Herrero, P.; Fogler, M. M.; Basov, D. N. Science 2014, 343 (6175), 1125−1129. (38) Dean, C. R.; Young, A. F.; Meric, I.; Lee, C.; Wang, L.; Sorgenfrei, S.; Watanabe, K.; Taniguchi, T.; Kim, P.; Shepard, K. L.; Hone, J. Nat. Nano 2010, 5 (10), 722−726. (39) Hwang, E. H.; Adam, S.; Das Sarma, S. Phys. Rev. Lett. 2007, 98 (18), 186806. (40) Yariv, A. Quantum Electronics; John Wiley & Sons: New York, 1975. (41) Blanco-Redondo, A.; Sarriugarte, P.; Garcia-Adeva, A.; Zubia, J.; Hillenbrand, R. Appl. Phys. Lett. 2014, 104 (1), 011105. (42) Wang, B.; Zhang, X.; Yuan, X.; Teng, J. Appl. Phys. Lett. 2012, 100 (13), 131111.

METHODS The calculations have been performed with the help of the finite elements methods using Comsol software. In order to simulate the propagation of the polaritonic mode, the profile of the mode in the analytical form has been set in the boundary conditions for the left boundary of the rectangular domain. The graphene layer has been modeled as a surface current in the boundary conditions. The domain has been discretized by using an inhomogeneous free-triangulat mesh with the maximal element size being less than 5% of the local spacial oscillation period of the field. The convergence has been assured both for electromagnetic fields and power fluxes (see Supporting Information).



ASSOCIATED CONTENT

S Supporting Information *

Description of additional details of the calculations. This material is available free of charge via the Internet at http:// pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*R. Hillenbrand. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was financially supported by the ERC Starting Grant no. 258461 (TERATOMO), the EC under Graphene Flagship (contract no. CNECTICT- 604391), and the Spanish Ministry of Economy and Competitiveness (Projects MAT2011-28581C02 and MAT2012-36580).



REFERENCES

(1) Hanson, G. W. J. Appl. Phys. 2008, 103 (6), 064302. (2) Vafek, O. Phys. Rev. Lett. 2006, 97 (26), 266406. (3) Wunsch, B.; Stauber, T.; Sols, F.; Guinea, F. New J. Phys. 2006, 8 (12), 318. (4) Jablan, M.; Buljan, H.; Soljačić, M. Phys. Rev. B 2009, 80 (24), 245435. (5) Koppens, F. H. L.; Chang, D. E.; García de Abajo, F. J. Nano Lett. 2011, 11 (8), 3370−3377. (6) Nikitin, A. Y.; Guinea, F.; Garcia-Vidal, F. J.; Martin-Moreno, L. Phys. Rev. B 2011, 84 (19), 195446. (7) Vakil, A.; Engheta, N. Science 2011, 332 (6035), 1291−1294. (8) Chen, J.; Badioli, M.; Alonso-Gonzalez, P.; Thongrattanasiri, S.; Huth, F.; Osmond, J.; Spasenovic, M.; Centeno, A.; Pesquera, A.; Godignon, P.; Zurutuza Elorza, A.; Camara, N.; de Abajo, F. J. G.; Hillenbrand, R.; Koppens, F. H. L. Nature 2012, 487 (7405), 77−81. (9) Fei, Z.; Rodin, A. S.; Andreev, G. O.; Bao, W.; McLeod, A. S.; Wagner, M.; Zhang, L. M.; Zhao, Z.; Thiemens, M.; Dominguez, G.; Fogler, M. M.; Neto, A. H. C.; Lau, C. N.; Keilmann, F.; Basov, D. N. Nature 2012, 487 (7405), 82−85. (10) Grigorenko, A. N.; Polini, M.; Novoselov, K. S. Nat. Photon 2012, 6 (11), 749−758. (11) Fang, Z.; Wang, Y.; Schlather, A. E.; Liu, Z.; Ajayan, P. M.; García de Abajo, F. J.; Nordlander, P.; Zhu, X.; Halas, N. J. Nano Lett. 2013, 14 (1), 299−304. (12) Fang, Z.; Thongrattanasiri, S.; Schlather, A.; Liu, Z.; Ma, L.; Wang, Y.; Ajayan, P. M.; Nordlander, P.; Halas, N. J.; García de Abajo, F. J. ACS Nano 2013, 7 (3), 2388−2395. (13) Low, T.; Avouris, P. ACS Nano 2014, 8 (2), 1086−1101. (14) Gao, W.; Shu, J.; Qiu, C.; Xu, Q. ACS Nano 2012, 6 (9), 7806− 7813. 2901

dx.doi.org/10.1021/nl500943r | Nano Lett. 2014, 14, 2896−2901